主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2024, Vol. 32 ›› Issue (4): 271-278.doi: 10.16381/j.cnki.issn1003-207x.2021.1658cstr: 32146.14.j.cnki.issn1003-207x.2021.1658

• • 上一篇    下一篇

基于PLS-Aenet的多工序制造过程关键质量特性识别

王宁1,3,田淑珂1,刘玉敏1,赵哲耘2,3()   

  1. 1.郑州大学商学院, 河南 郑州 450001
    2.郑州大学规划与学科建设部, 河南 郑州 450001
    3.郑州大学国际质量发展研究院, 河南 郑州 450001
  • 收稿日期:2021-08-20 修回日期:2022-03-03 出版日期:2024-04-25 发布日期:2024-04-25
  • 通讯作者: 赵哲耘 E-mail:zhaozheyun@zzu.edu.cn
  • 基金资助:
    国家自然科学基金重点项目(U1904211);国家社会科学基金项目(20BTJ059);河南省高等学校青年骨干教师培养计划(2021GGJS006);河南省高校哲学社会科学创新人才支持计划(2023-CXRC-19);郑州大学精尖学科支持项目(XKLMJX202201);郑州大学人文社会科学优秀青年科研团队(2023-QNTD-01)

Identification of Key Quality Characteristics in Multistage Manufacturing Process Based on PLS-Aenet

Ning Wang1,3,Shuke Tian1,Yumin Liu1,Zheyun Zhao2,3()   

  1. 1.School of Business, Zhengzhou University, Zhengzhou 450001, China
    2.Department of Planning and Discipline Development, Zhengzhou University, Zhengzhou 450001, China
    3.International Institution for Quality Development, Zhengzhou University, Zhengzhou 450001, China
  • Received:2021-08-20 Revised:2022-03-03 Online:2024-04-25 Published:2024-04-25
  • Contact: Zheyun Zhao E-mail:zhaozheyun@zzu.edu.cn

摘要:

复杂多工序制造过程包含较多质量特性,且质量特性具有多重共线性、高维度及群组效应等特点,这些特点使得关键质量特性识别较为困难。为解决此问题,本文运用偏最小二乘回归改进自适应弹性网(Aenet)方法并结合状态空间模型进行建模分析,从而进行复杂多工序制造过程关键质量特性识别,并给出了基于PLS-Aenet方法的关键质量特性识别具体步骤,最后通过仿真分析对比PLS-Aenet方法与岭回归、Lasso、弹性网以及PLS-Lasso等变量选择方法在复杂多工序制造过程关键质量特性识别中的适用性与有效性。仿真与实例结果表明,相比于岭回归等变量选择方法,本文所提改进的PLS-Aenet方法在关键质量特性间具有较强相关性时识别效果更佳,特别是对具有群组效应的关键质量特性具有更优的识别精确性和完整性。

关键词: 多工序制造过程, 复杂产品, 状态空间模型, PLS-Aenet, 关键质量特性

Abstract:

The complex multistage manufacturing process contains many quality characteristics which are characterized by high dimensional multicollinearity and group effect. In the actual production process, subjected to the cost and technology constraints, all quality characteristics cannot be monitored. How to identify the key variables from a lot of quality characteristics with high dimensional multicollinearity and group effect has become increasingly necessary. To solve this problem, The Partial Least Squares- Adaptive Elastic Net (PLS-Aenet) method is applied to model and analyze the complex multistage manufacturing process and identify the key quality characteristics of the process. The PLS-Aenet method is used and the state space idea is integrated to model and identify the key quality characteristics of complex multistage manufacturing processes, and the specific steps are given to identify the key quality characteristics based on the PLS-Aenet Net method. Finally, the applicability and effectiveness of the PLS-Aenet Net method and the variable selection method such as Ridge Regression、Lasso、PLS-Lasso、Elastic Net method in the identification of key quality characteristics of the complex multistage manufacturing process are compared through simulation analysis. The simulation and real case results show that the PLS-Aenet method achieves a more effective performance when the key quality characteristics are strongly correlated with each other.

Key words: multistage manufacturing process, complex product, state space model, PLS-Aenet, key quality characteristics

中图分类号: